Multidimensional Response Surface Methodology for the Development of a Gene Editing Protocol for p67phox-Deficient Chronic Granulomatous Disease

Abstract

Replacing a faulty gene with a correct copy has become a viable therapeutic option as a result of recent progress in gene editing protocols. Targeted integration of therapeutic genes in hematopoietic stem cells has been achieved for multiple genes using Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)/Cas9 system and Adeno-Associated Virus (AAV) to carry a donor template. Although this is a promising strategy to correct genetic blood disorders, it is associated with toxicity and loss of function in CD34⁺ hematopoietic stem and progenitor cells, which has hampered clinical application. Balancing the maximum achievable correction against deleterious effects on the cells is critical. However, multiple factors are known to contribute, and the optimization process is laborious and not always clearly defined. We have developed a flexible multidimensional Response Surface Methodology approach for optimization of gene correction. Using this approach, we could rapidly investigate and select editing conditions for CD34⁺ cells with the best possible balance between correction and cell/colony-forming unit (CFU) loss in a parsimonious one-shot experiment. This method revealed that using relatively low doses of AAV2/6 and CRISPR/Cas9 ribonucleoprotein complex, we can preserve the fitness of CD34⁺ cells and, at the same time, achieve high levels of targeted gene insertion. We then used these optimized editing conditions for the correction of p67^phox-deficient chronic granulomatous disease (CGD), an autosomal recessive disorder of blood phagocytic cells resulting in severe recurrent bacterial and fungal infections and achieved rescue of p67^phox expression and functional correction of CD34⁺-derived neutrophils from a CGD patient.

Keywords: AAV; CGD; gene editing; homology-directed repair; response surface methodology.

Figure 1. Development of a gene editing platform for the treatment of p67-CGD.

a) The AAV6 co-p67^phox, used as an HDR donor, comprises 400 bp left (HAL) and right (HAR) homology arms, a codon-optimised p67 ^phox cDNA exon 1-15 (co-p67 ^phox) and a BGH poly-A sequence (pA). A representation of the NCF2 locus is shown, including the sgRNA sequence (sgRNAT89) cutting in exon 1 (15 nucleotides before the ATG), and the 3 sgRNAs used to knock out the genomic NCF2 by mutating exon 3. Arrows show the approximate cut site of sgRNAs. b) Sequence of wildtype and knockout NCF2 Exon 3. Sequencing confirmed an inactivating frameshift mutation in Exon 3, in which a T-insertion leads to a premature stop codon. c-d) Expression of p67 ^phox by wildtype PLB-985s, a knockout clone (KO) and mono and bi-allelic corrected clones, with and without differentiation into neutrophils as assessed by (c) Western blot and (d) Flow cytometry, with Median Fluorescence Intensities of three technical replicates (±SD) shown for each condition (e).

Figure 2. Design of the RSM experiment showing values selected for each level of each factor.

The experiment follows a central composite design (CCD) (a). Levels are tested in three types of combinations: all central points, Low/High points (“Cube” points), and Central/Extreme points (“Star” points). The left panel shows a RSM design for three factors. 17 combinations are tested in this design: 3 repeated centre points (CCC), 8 “Cube” points (L/L/L, L/L/H, L/H/L, H/L/L, H/H/L, H/L/H, L/H/H, H/H/H), and 6 “Star” points (C/C/XL, C/C/XH, C/XL/C, C/XH/C, XL/C/C, XH/C/C,). Values are chosen such that these combinations are rotatable with respect to the central set of conditions. Two overlapping CCDs are used (b), with the Low MOI conditions for the first design identical to the High MOI conditions for the second (shown in bold italics), giving 9 MOI levels in total. With the overlapping design, 30 combinations are tested (17 + 17 – 4 overlapping MOI “Cube” point conditions).

Figure 3. RSM for the optimization of parameters for the gene editing of p67-CGD

a) The raw data for all three replicates is shown for i) fractional copy number, ii) the fractional reduction in total cell number at Day 5 compared to an electroporation-only control and iii) the fractional reduction in colony-forming units compared to an electroporation-only control. b) The fitting of the regression model of fractional copy number. i) shows residuals vs fits indicating good fit except at the lowest observed copy number values. S is a measure of fit given by the average distance of an observed data point from the models’ prediction of that point. 10-fold S indicates the predictive power of the model, with a value close to S indicating good predictive capacity. R² cannot be used as an indicator of fit as the intercept for the model is set to 0. ii) The terms included in the model. Terms with an α < 0.15 were included in the stepwise model generation. Terms below α = 0.05 (p < 0.05) are statistically significant and are ranked by standardised effect size. Higher-order and interacting terms are included as described in the methods. iii) The response surface of the model for copy number for Donors 1-3 (Cas9 kept constant at 8 μg/10⁶ cells). c) and (d) are as for (b) but for Cell loss and CFU Loss respectively. e) Desirability functions for maximisation of copy number and minimisation of Cell and CFU loss are shown for i) each donor/replicate and ii) as a normalised average of the 3 donor/replicates with iii) standard deviation.

Figure 4. Predictions and relationship to viability.

a) Predicted mean values for each donor/replicate are shown with error bars indicating confidence interval (95% confident true mean within this range) and predictive interval (95% confident new observations will fall in this range), for Copy number (i), Fractional cell loss (ii) and Fractional CFU loss (iii). b) Correlations between i) Viability and CFU loss, ii) Viability and Cell Loss, and iii) CFU loss and Cell loss are shown.

Figure 5. Testing the optimality of AAV6 co-p67^phox at MOI 2800.

HSPCs were edited as before with MOIs of 53 (low), 2800 (optimum), and 13325 (high). Copy number over time is shown in panel a) and the projected number of corrected cells at each timepoint in panel b) (n=3; mean ± SD. *: p<0.05, ****: p<0.0001 (2-way ANOVA & Post-Hoc Tukey test). c) Western blotting showing expression of p21 in response to editing at Day 1 (n=2).

Figure 6. Functional correction of patient-derived p67^phox-deficient cells.

a-b) Neutrophil differentiation (as assessed by cd11b expression) and p67^phox expression of wildtype (WT) and p67^phox-deficient (p67-) primary CD34+ cells by flow cytometry (a) and western blotting (b). c) Dihydrorhodamine (DHR) assay (shown is the percentage of rhodamine 123 +ve out of cd11b +ve cells) and d) Nitro Blue Tetrazolium (NBT) assay of WT, patient cells and gene edited cells. Scale bar = 50 μm.